Ask Your Question

Revision history [back]

Because in boosting, you want to find new samples that do not get correctly classified by your previous weak stages. This means that all your artificially generated samples are already perfectly seperated by your 5 stage classifier and thus it needs new data if you want to continue training.